Local validation of global biomass maps

Abstract

Listen

Translate article

Tropical countries without extensive ground sampling programs often use the gain-loss approach for greenhouse gas inventories. With this approach emissions are estimated as the products of estimates of areas of land use change characterized as activity data and estimates of emissions per unit area characterized as emission factors. For the special case of complete deforestation, the emission factor for the aboveground, live tree, biomass (AGB) pool can be estimated as mean biomass per unit area prior to the deforestation which, in turn, can be estimated using a single-date ground sample or a single-date biomass map.Multiple biomass maps at continental and global scales that can be considered for estimation of emissions factors have been forthcoming in recent years. However, regardless of the source of biomass data, estimates of emissions factors must still comply with the IPCC good practice guidelines for accuracy and uncertainty. For emission factors based on a biomass map, compliance requires validation of the map where validation means that independent reference data are used to ensure that the map is free from systematic error.The primary objective of the study was to illustrate methods for validating global AGB maps by assessing their conformity to reference data at the spatial scale of a sub-global validation site. For this study, validation entailed assessing the accuracy of map-based estimates of mean biomass per unit area for a sub-global, approximate 7600-km2 validation site in Minnesota, United States of America (USA). The accuracy assessment was in the form of a test of the hypothesis of whether the map-based estimate is or is not equivalent to an estimate based on independent reference data obtained from either a ground sample or a local biomass map of greater quality than the global map. Design-based, model-based, and hybrid inferential methods were used to conduct the test.The primary conclusions were threefold. First, no map exhibited systematic error; second, the model-based inferential approach produced the smallest standard errors; and third, operationally, differences between validation site estimates of mean biomass per unit area for two global biomass maps and estimates based on reference data were not statistically significantly different.